How to Hire a Digital Marketing Agency in 2026: 10 Questions
Vet agencies with 10 must-ask questions on AI, SEO, and growth. A practical checklist for founders and CMOs to protect budget and pick the right partner.
A founder I know burned $84,000 last year on an agency that promised "full-funnel growth." Six months in, the dashboard looked busy: impressions up, engagement up, a weekly report nobody read. Pipeline? Flat. When she finally asked which campaigns had produced closed revenue, the account manager needed two weeks to answer. That delay was the tell. A team that owns your numbers can pull them in an afternoon.
The agency market has gotten harder to read, not easier. More shops now lead with AI in the pitch deck, and plenty of them mean something vague by it. Meanwhile the fundamentals that actually move CAC and LTV have not changed much. What changed is how easy it is to hide weak work behind automated reporting and confident language.
So here is the checklist we use when we evaluate partners, or when clients ask us to sit in on their [selection process](/dgtg/blog/how-to-choose-a-digital-marketing-agency-tips-for-founders-cmos). Ten questions, roughly in the order you should ask them.
## 1. What does "AI" actually mean in your workflow?
This is the first question because it filters fast. "We use AI" is not an answer. Ask them to name the tools and the step in the process where each one sits. Do they use it for creative variants, media buying bid strategies, [keyword clustering](/tech/blog/ai-minimalism-how-to-simplify-your-tech-life-in-an-ai-world), or reporting summaries? Who reviews the output before it ships?
An agency doing this well will talk about guardrails: brand voice checks, human review on anything client-facing, and a clear line between draft and final. One that is bluffing will pivot to a case study.
## 2. How do you measure success in the first 90 days?
Beware anyone who promises ROAS in month one. Paid search can show signal quickly, but SEO, content, and lifecycle work need runway. A credible answer separates leading indicators (CTR, qualified demo requests, branded search volume) from lagging ones (CAC, payback period, LTV by cohort).
Ask them to put the 90-day targets in the contract. If they resist, that is your answer.
## 3. Which parts of this should we keep in-house?
This is the question most founders skip, and it is the one that saves the most money. Good agencies will tell you what not to outsource. In our experience, the split usually looks like this:
| Keep in-house | Hire out |
|---|---|
| Positioning and messaging strategy | Paid media execution and bid management |
| Customer interviews and sales feedback loops | Technical SEO audits and fixes |
| Pricing and offer design | Content production at scale |
| CRM data ownership | Reporting infrastructure |
An agency that wants to own your positioning is usually selling you something you cannot easily replace. An agency that wants to own your ad accounts is normal, as long as you hold admin access.
## 4. Show me a client you fired, or one that fired you.
Everyone shows the wins. The interesting data is in the losses. Ask what went wrong, how fast they caught it, and what they changed. You are listening for specifics: a client that outgrew the retainer, a channel that stopped working, a mismatch in expectations about speed.
Vague answers here are a red flag. Agencies that have been around long enough have stories. The good ones tell them without flinching.
## 5. Who actually does the work?
The people in the pitch are often not the people on your account. Ask for names, roles, and the percentage of time each will spend on your business. Then ask what happens when your strategist goes on leave or leaves the company.
Get this in writing. Turnover is the single most common reason agency relationships sour, and it is rarely disclosed upfront.
## 6. How do you handle attribution?
This one separates the operators from the reporters. Ask how they credit a sale that touched three channels before converting. Listen for whether they use a single model (last click, first click) or a blended view with incrementality testing.
If they cannot explain their attribution model in plain language, they cannot optimize against it. And if they claim perfect attribution, they are either new or lying. Nobody has it.
## 7. What is your reporting cadence, and what is in it?
Weekly calls are not the same as weekly insight. Ask what decisions the report is meant to drive. A good report answers three questions: what changed, why, and what we are doing about it next.
Push back on vanity metrics. Impressions and follower counts belong in an appendix, not a headline. If a metric does not connect to revenue or a leading indicator of it, it should not be on page one.
## 8. How do you price, and what is not included?
Retainers, performance fees, media spend minimums, and pass-through costs all vary. Ask for a written scope with exclusions. Common surprises: creative production billed hourly, ad platform fees, and strategy work that quietly falls outside the retainer.
Also ask about contract length and exit terms. Month-to-month is rare but not unheard of. A 12-month lock-in with no performance clause is a bet you should not take in your first engagement.
## 9. What does your tech stack look like, and who owns the accounts?
You should own your ad accounts, analytics property, and CRM. The agency should have access, not ownership. This matters more than people think. Switching agencies is painful; switching agencies and rebuilding your data history is worse.
Ask which tools they use for SEO (Ahrefs, Semrush, Screaming Frog), paid media, and reporting (Looker Studio, Triple Whale, or a custom warehouse). Tool choice tells you something about how they think.
## 10. What would make you tell us to stop?
This is the honesty test. A good partner will name conditions: if the unit economics do not support paid acquisition, if the offer needs work before scaling spend, if the market is too small for the budget. Anyone who says "we can always find a way" is selling, not advising.
## Our take
If you are evaluating agencies this year, weight these three things above everything else: transparency on attribution, clarity on who does the work, and a willingness to tell you what to keep in-house.
On tooling, we lean toward agencies that run their own reporting warehouse (BigQuery or similar) rather than relying only on platform dashboards. Platform-reported ROAS is optimistic by default. On SEO specifically, look for teams that treat AI-generated content as a drafting tool with human editing, not a publishing strategy. Google's helpful content signals have been consistent on this for two years running.
If you want a shortlist to benchmark against, the shops that consistently show up well in our conversations are those with published case studies that include actual numbers (CAC, payback, revenue lift) rather than logos and adjectives. Demand numbers. The ones who have them will be happy to share.
The rest of the evaluation is fit: communication style, speed of response, and whether they push back when you are wrong. You are hiring judgment, not just execution.
## FAQ
### How long should we give a new agency before judging results?
Ninety days for paid channels, six months for SEO and content. Set leading-indicator targets for the first quarter and revenue targets for the second.
### Should we pay for performance or a retainer?
Retainers align better for most B2B work because the effort is front-loaded. Hybrid models (base plus bonus on agreed metrics) work when the metric is clean and attributable.
### What is the biggest red flag in agency selection?
No named team, no attribution model, and no willingness to define what failure looks like. Any one of those alone is worth a second look.
Frequently asked questions
How long should we give a new agency before judging results?
Ninety days for paid channels, six months for SEO and content. Set leading-indicator targets for the first quarter and revenue targets for the second.
Should we pay for performance or a retainer?
Retainers align better for most B2B work because the effort is front-loaded. Hybrid models (base plus bonus on agreed metrics) work when the metric is clean and attributable.
What is the biggest red flag in agency selection?
No named team, no attribution model, and no willingness to define what failure looks like. Any one of those alone is worth a second look.
DGTG
